Atlas

Benchmarks

← All benchmarks

BioMysteryBench - Human Difficult (original problem set)

Science

BioMysteryBench Human Difficult as reported before July 2026, before Anthropic removed 6 problems from this subset on external feedback. Scores on the original set are not comparable to the revised subset. Diagnostic because the set was superseded.

Top models (higher is better)

ModelScore
Claude Fable 546.1
Loading Atlas data…